[Standards-JIG] XMPP URIs was: Two questions regardingJEP-0124HTTP Binding
mridul at sun.com
Mon Nov 28 10:02:29 UTC 2005
Hi Ian , Peter,
Thanks for the clarifications.
A way to specify the route host is required - another example (other
than what has been mentioned before) is the XMPP server returning a
streamerror with see-other-host.
Peter Saint-Andre wrote:
> Ian Paterson wrote:
>> Peter is proposing that the 'route' attribute should be a simple
>> "host[:port]" value. Other possibilities *might* be "xmpp:host[:port]"
>> (compatible with existing JEP-0124 spec) or just "xmpp:host". The latter
>> is compatible with XMPP URI/IRIs ('route' is after all an XMPP resource
>> identifier), but perhaps it is not ideal, see
> OK, so we have three options:
> 1. host[:port]
> Pro: specifies everything we need
> Con: not backwards-compatible with what we have now
> doesn't support (future?) non-XMPP connections
> 2. xmpp:host[:port]
> Pro: backwards-compatible with what we have now
> Con: looks like an XMPP URI but isn't (confusing?)
Option 1 is sufficient for now ... but potentially limiting.
Using "xmpp:host:port" or "xmpp://host:port" (or variations thereof)
would take care of potential future requirements from what we can
envision now ...
Since there are already implementations of clients and
servers/connection managers out there - the changes made should be , if
possible , as backwardly compatible as possible.
> 3. xmpp:host
> Pro: ?
> Con: doesn't support ports
> not backwards-compatible with what we have now
> I think (3) is a non-starter. I don't like the confusing aspect of (2)
> and personally I doubt that this spec will ever be used for non-XMPP
> connections, so I prefer (1). Have I missed any pros and cons?
This is definitely not very optimal - port can be a requirement to
disambiguate the actual server to connect to.
More information about the Standards